The linguistic fact is best characterised by processes: production process, reception process, acquisition, and linguistic change. In trying to understand these processes, one bears a handicap when adopting a static, 'grammar'-like approach, or if postulating a 'language', even an individual or internal language: from a static object it is too difficult to then backtrack to the processes. Facing the processes right away would seem a more useful route.
The primary object is thus shifted: it is no longer a language, it is the speaking subject as a processing entity; the speaking subject knows how to, rather than knows that: a child does not learn a language, he just learns how to speak.

Two interdependent postulations are made:

(1) a static model, consisting of exemplars (later episodes), with paradigmatic links between them (structure mappings, or analogies if you like); there is no abstraction, no rule; linked in this way, the exemplars constitute a complex mesh: a 'plexus'.

(2) a processing model, that computes an act – e.g. a reception act – against a plexus. The processing model is an abductive solver; it takes advantage of the proximities embodied by the paradigmatic links; it licences novelties using the most accessible exemplars, and thus accounts for the infiniteness of linguistic productivity. This model renders regularisation effects without postulating rules, and categorisation effects without postulating categories.

This is not simply programmatic. One such model is today effective for most of syntax and morphology; it is precise and demonstrable while few propositions are today. There is only one secret: precisely to abstain positing a language, which would compromise the enterprise. This also holds for linguistic change.

In this formal perimeter the model is restricted to formal terms (phonological or orthographical). However, it already encompasses a fist-level stance about how the know-how of a bilingual subject is organised. In order to extend the model to 'semantics', private terms complement the formal terms. So the exemplars, so far restricted to linguistic form, are complemented with a private part and thus become episodes. Meaning effects then obtain without its being necessary to posit 'significations' or 'meanings'.

Languages, spoken to some extent : European languages, Latin, Japanese.
Interest in Inuktitut, Classical Nahuatl, Yaqui, Swahili.